LINGUIST List 12.1683

Wed Jun 27 2001

Review: Culicover & Postal, Parasitic Gaps

Editor for this issue: Terence Langendoen <>

What follows is another discussion note contributed to our Book Discussion Forum. We expect these discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for discussion." (This means that the publisher has sent us a review copy.) Then contact Simin Karimi at or Terry Langendoen at


  1. sharbani, Review: Culicover & Postal, Parasitic Gaps

Message 1: Review: Culicover & Postal, Parasitic Gaps

Date: Thu, 28 Jun 2001 03:02:13 +0530
From: sharbani < net">sharbevsnl. net>
Subject: Review: Culicover & Postal, Parasitic Gaps

Culicover, Peter W. & Postal, Paul M., ed. (2001) Parasitic Gaps.
MIT Press, hardback ISBN: 0-262-03284-8, ix+447pp., $55. 00
(Current studies in Linguistics 35).

Reviewed by Sharbani Banerji, Centre for Applied Linguistics and
Translation Studies, University Of Hyderabad, India.

A Parasitic Gap (P-gap) is a construction of the following kind:

a) which articles did John file t without reading pg?

The first gap marked 't' is called the true gap, from where extractions
normally occur, while the second gap marked 'pg' is the parasitic gap
because it is dependent on the existence of another gap, the true gap.

A book devoted entirely to Parasitic gaps is first of it's kind, and is
going to be the most valuable book for all the those
students, researchers, professors, scholars and professionals who are
pursuing the topic of Parasitic Gaps in one form or the other. It is
as valuable to a student who has just initiated the study on the subject
of Parasitic Gaps, as to a scholar who is deeply interested in the
subject and is working on it's finer intricacies. The editors have
attempted to bring all the important and relevant literature on
Parasitic Gaps under one roof. As a result, it has an enormous amount
of data in it. Though the major work on P-gaps till date has been done on
English, this book has works on other languages too, viz. , Swedish,
Hungarian, Moroccan Arabic, French, Standard German, Dutch, Persian,
Hindi, Hebrew, etc.

The work on one hand brings out the need to work on this topic in
other languages, so that the acclaimed theories can be verified, and on
the other hand it makes that work so much simpler by literally handing
over all the core data on a platter to a researcher.

For a student, Peter W. Culicover's article, the first chapter in the
book, summarizes very systematically, and in a simple, readable style,
the entire development in the field of P-gaps until now, starting from
Engdahl's 1983 seminal work on the subject.

The volume has thirteen papers by different authors, some reprinted here,
and some printed for the first time. Of the two editors, Peter
W. Culicover has one paper, and Paul M. Postal has three papers, one of
which is his 1994a paper, reprinted here, while the other two are new.
Elizabet Engdahl has two papers, one of which is her 1983 paper, the
seminal work on Parasitic gaps. The second one is printed here for the
first time. Katalin E. Kiss's work is an updated version of her 1985
paper. Similarly, Christopher Kennedy's work is his 1997 paper. All other
works are new and have been published here for the first time.

The book is divided into four major sections.

Section I is historical in nature. It has three papers. The first paper
is by Peter W. Culicover, which gives the historical background of
P-gaps, relating to it's properties and theories. The second paper is the
reprinted version of Engdahl 1983, and the third is the updated version
of Kiss 1985, which focuses on the role of case in P-gap distribution,
and works on Hungarian and English data.

Section II centers on the identifying characteristics of P-gaps. It
contains four new papers.

Engdahl's paper: She works mainly on Swedish data to deal with the
relation between P-gaps and category restrictions, anaphoric elements,
referentiality and island facts.

Ouhalla's paper focuses on the P-gaps in Moroccan Arabic and their
relation to resumptive pronouns.

Levine, Hukari and Calagno's paper dispute the claim that P-gaps are of
the category NP and have inherently pronominal properties.

Postal's long paper considers the pronominal character of P-gaps also in
the licensing gap.

Section III contains two papers addressing the boundaries of the P-gap

Postal's 1994a paper is reprinted here. It makes a distinction between
Parasitic and Pseudo Parasitic gaps, analyzing the Pseudo P-gaps as a
case of Right Node Raising.

Kathol's new paper argues that Standard German lacks true P-gaps.
Instead, they are more like the Pseudo P-gaps of Postal 1994a.

Section IV contains four papers, three published here for the first
time, dealing with factors that limit the occurrence of genuine P-gaps.

Tellier's paper compares French and English P-gaps and relates the
differences to the distinction in the agreement systems of the two

Munn's paper treats the relation between P-gaps and logical concepts
such as individual variable as well as the link between P-gaps and
coordination and pronouns.

This section also contains Kennedy's 1997 paper which argues that
contrary to an earlier claim, a certain class of cases do not involve
P-gaps "inside" elided VPs, or indeed any P-gaps at all, but are rather
ordinary pronouns as an instance of what Fiengo & May 1994 call "vehicle

Postal in his paper extends Kennedy's argumentation and argues that
although Kennedy's claim is essentially correct for the class of cases
he treated, application of the same and related argumentation to a
different range of data supports the claim that English has P-gaps
"inside" of verbal ellipsis structures.




In this Chapter Culicover traces the history of the development of the
theories of Parasitic Gaps from the time of Engdahl 1983. He points out
that though the phenomenon of P-gaps had been noted previously by Ross
1967, Bresnan 1977, Taraldsen 1981, the real work on P-gaps was initiated
by the publication of Engdahl 1983.

As he traces the history, he brings out the major issues concerning
P-gaps. Though there have been a broad agreement on these issues, which
he calls Current Consensus Position (CCP), challenges have been posed to
these CCPs from time to time. First he identifies the six CCPs, and next
he discusses the challenges to these CCPs.

The six CCPs are:

CCP1: The antecedent of a P-gap must be in an A'-position.

CCP2:A P-gap is licensed only at S-structure.

CCP3: The antecedent of a P-gap must be an NP.

CCP4: The true gap cannot c-command the P-gap.

CCP5:The P-gap is in a chain with the antecedent of the true gap.

CCP6: Anti-c-command is a consequence of Condition C of the Binding

To start with, Culicover discusses Engdahl's paper in detail,
because it identifies majority of the questions relating to P-gaps.

Kayne 1983 observed that P-gaps within islands embedded in other
islands produce ungrammaticality comparable to movement. If there is no
movement involved in P-gaps, and hence ungrammatical cases cannot be
ruled out by Subjacency, nor by ECP, since P-gaps are properly governed,
then how does one account for such cases?

Kayne therefore proposes an extension of the ECP such that the entire set
of g-projections of all the gaps in a tree must constitute a subtree that
is locally c-commanded by the antecedent. In Kayne's analysis, the
antecedent of the true gap forms a chain with the P-gap that is subject
to the same locality conditions that govern all chains. Hence CCP5.
Chomsky 1986b, in Barriers explains the phenomenon by proposing that
there is a null operator that binds the P-gap. The P-gap on this analysis
is the trace of the null operator and the chain produced by movement of
the null operator is subject to the usual locality conditions on
movement. To relate the antecedent of the true gap and the P-gap, Chomsky
proposes that the chain of the true gap and that of the P-gap are
"composed" in the sense that the P-gap is c-commanded and A-bound by the
true gap, leading to Condition C violation. Similarly, an antecedent in
A-position A-binds the P-gap and produces Condition C violation. Hence



The CCP is that the antecedent of the P-gap must be an NP. The
restriction may be related to the pronominal character of the gap, as
pointed out by Cinque (1990:115) . Though Italian and English support this
generalization, Engdahl 1983 cites Swedish examples in which the
antecedent of the P-gap is not an NP. Similarly, Levine, Hukari and
Calagno in Chapter 6 make a stronger claim that non-NP P-gap may occur
even when there is no suitable overt proform in the language. Again, CP
antecedents can also license P-gaps.

Tellier 1991 shows that in French, phrases with "dont" (a PP plausibly),
can be antecedents of a P-gap.


CCP holds that the antecedent must be in an A'-position. But
observations about apparent P-gap constructions involving clitic movement
in the Romance languages, Heavy NP Shift in English, and scrambling in
languages such as German, Hindi and Persian have raised the possibility
that the antecedent of the P-gap could also be in an A-position. On the
otherhand, it is possible that such apparent A-movements are actually
A'-movements, and a number of researchers have taken this position.

CCP holds that a P-gap forms a chain with some antecedent and hence is a
trace. The literature also contains arguments that it is pronominal. There
are two variants of both approaches.

Trace analysis:

i) t1, the empty category, forms a chain with the antecedent
of the true gap.

ii) t2, the empty category, forms a chain with the null operator
which in turn is linked to the chain that contains the antecedent of the
true gap.

Pronoun analysis:

i) P1, the P-gap, is an empty pronoun.

ii) P2, the P-gap, is an anaphor, related to a reflexive or trace
of the NP movement.

P-gaps as TRACES:

i) This was originally proposed by Chomsky 1982, but was undermined in
two ways. First, Kayne 1983 showed that P-gap shows Subjacency
effects. Second, Brody 1984 showed that Chomsky's theory of the
contextual definition of empty categories was in part factually
incorrect, and in part followed from the independent grammatical
principles such as ECP and the theta criterion.

ii) Frampton 1990:Traces are base generated and the Subjacency is a
condition on the well-formedness of chains. On his analysis, there are two
chains, one containing the true gap and the other containing the
P-gap. The head of the P-gap chain is replaced by a trace, allowing this
chain to be linked to the higher chain. Movement of the antecedent of the
true gap introduces an intermediate trace that locally c-commands the
trace that heads the P-gap chain. As a consequence, a chain is formed
between the P-gap and the antecedent.


i) Cinque argues that there is a class of apparent extractions that are
the consequence not of movement but of binding of an empty pronominal by
an operator. In certain cases where extraction is impossible, which Cinque
calls "weak islands", the empty pronominal can be bound. This differece
makes it appear that only NPs can be extracted from such islands. P-gaps
seem to display the same behaviour with respect to weak islands, a result
that follows if P-gap is a pronominal. Levine, Hukari and Calagno in
Chapter 6 criticize Cinque's proposal.

Postal gives additional evidence that a P-gap is a pronominal. He argues
that P-gaps do not appear in contexts that are anti-pronominal, e. g. in
"there constructions". He presents evidence that shows that the true gap
may not occur in a class of contexts that exclude weak pronominals.
Culicover also cites Hornstein 1995:172, to prove that a P-gap is an
empty pronominal.

Additional support for pronominal analysis is given by Ouhalla (chapter
5) who shows that the behavior of P-gaps in Moroccan Arabic can be given
an elegant explanation if they are taken to be null resumptive pronouns.

ii) is proposed by Bordelois 1986:12 who observes that P-gaps can appear
in questions. Her proposal is that the P-gap is actually an empty anaphor.

Other views about the character of the gap:
Lasnik & Stowell 1991 address the fact that P-gap constructions in
general do not appear to yield weak cross over (WCO) effects in violation
of the Bijection Principle of Koopman & Sportiche 1982. According to
them, the WCO effect occurs only when the operator is a "true Quantifier
Phrase". They argue that the trace of a non-QP null operator is not a
variable, but a null epithet. In the P-gap construction, the null epithet
is an r-expression, and hence a Condition C violation is produced when it
is c-commanded by a true gap.


On the assumption that there are two chains in the P-gap construction
with a common head, it would follow that the true gap and the P-gap would
have to have the same case, the case of the head. However, if we take the
position that the P-gap is a pronominal, then it does not form a chain
with the antecedent of the true gap, and hence there is a possibility
that its case may be different from that of the true gap.

The relevance of case compatibility was first noted by Kiss 1985,
reprinted in Chapter 3. In Hungarian, the cases of the P-gap and the true
gap must be identical, in the sense that the overt case marking on the
antecedent must be compatible with the chain that it forms with the true
gap and the chain that it forms with the P-gap.

Similar case compatibility facts were pointed out by Franks 1992 in
Slavic languages, particularly in Polish and Russian.


The original observation that a subject cannot license a P-gap if it
c-commands it is due to Engdahl 1983. However, counterexamples exist. So,
the relevant property would indeed be that of c-command.

Accounting for the condition of anti-c-command:

A natural approach is to relate it to binding theory, which in it's
classical form crucially invokes the relation of c-command. Which of the
conditions is involved is thus tied to the argument for the character of
the gap.

Generalized Phrase Structure Grammar (GPSG) account.

Aoun and Clark 1985 propose that the non-overt operator in P-gap
constructions is treated as an A'-anaphor, which is subject to a
generalized binding theory.

In the Connectedness account of Kayne 1983 and Longobardi 1984, the
anti-c-command condition is a consequence of the definition of the
Connectedness condition.

Counterexamples to the anti-c-command condition:

Contreras 1984 observes that there is an apparent conflict between the
anti-c-command condition and the fact that direct objects can bind into
adjuncts to their right.

More serious counterexamples have been noted by Horvath 1992.


Wahba 1995 gives evidence that wh-movement and resumptive pronouns
license P-gaps in standard Arabic, and that wh-in-situ does as well in
Jeddah Arabic. That is a counterexample to the claim that P-gaps are
licensed only at S-structure, because the gap associated with the
wh-in-situ is present only at Logical Form (LF), if at all.


Bordelois 1985 shows that in Spanish, P-gaps occur only in untensed
adverbial clauses, whereas in English, untensed adverbials are in general
preferred. There are no P-gaps in Spanish in domains that have an overt
subject in an adjunct, in contrast to English, where though a subject is
dispreferred, is not ungrammatical. P-gaps in Spanish are illicit when
there is an adjunct to the main VP that is unselected, and intervenes
between the main VP and the adjunct containing the P-gap. Clitic climbing
shows the same distribution as parasitic gaps. Bordelois argues that the
domain restriction of P-gaps in Spanish would follow if P-gaps are
anaphors and therefore subject to Condition A of the binding theory,
implying P-gaps are anaphoric variables.

Garcia-Mayo and Kempchinsky 1993 propose that the fundamental
difference between English and Spanish P-gaps is that in English the null
operator that binds the P-gap is moved into an A'-position and binds a
trace, whereas in Spanish, the null operator is generated in initial
position and binds a pro.

Felix 1985 argues that P-gap constructions in German are subject to
the same principles that govern English P-gaps. Dutch seems to lack P-gaps
of the English type. Huybregts and van Riemdijk 1985 attribute this
difference between English and Dutch to a difference in the government
properties of prepositions that serve as subordinate conjunctions.

Tellier in Chapter 10 discusses three main differences between English
and French P-gaps, all of which have to do with the domain of the P-gap.
Unlike in English, French disallows P-gaps in adjuncts and relative
clauses. And French permits P-gaps in definite NPs, whereas English
does not. Tellier attributes the observed differences to differences in the
agreement properties of the functional categories C and D in the two

ATB extractions:

Grosu 1980 was the first to suggest that P-gap constructions are an
extension of across-the-board (ATB) extraction from a coordinate
structure. Many other scholars have worked in this direction in different

Munn 1992 assumes that P-gaps are produced by across-the-board (ATB)
movement. His approach differs from earlier approaches in that for him,
coordinate ATB constructions are a subtype of P-gap constructions.

CHAPTER 2:PARASITIC GAPS: (Elisabet Engdahl)

It is the reprinted version of Engdahl 1983, the work that for all
practical purposes initiated the work on Parasitic Gaps. According to
Culicover, Engdahl's seminal paper either directly identifies or
immediately suggests the key issues about P-gaps, which are:

i) The location of the antecedent A/A'-position

ii) The level at which the pg is licensed

iii) The character of the pg (trace or pronominal or otherwise)

iv) The anti-c-command condition

v) The obligatory/optional distinction

vi) The domain of the pg

vii) The character of the antecedent, e.g., whether the antecedent is
actually the antecedent of the true gap, or whether the antecedent is
some empty element that is linked to the antecedent of the true gap.

Engdahl works on Swedish data mainly. She talks about the distribution
of P-gaps, and gives the following "Accessibility hierarchy" for
occurences of P-gaps:

Manner adverb > temporal adverb > purpose clauses---untensed domains
> that/than clauses > when/because/conditional "if-clauses" > relative
clauses/indirect questions-----tensed domains.

She gives examples to justify her claim.

According to her, it is the actual presence of a real gap that licenses a
P-gap and not just the presence of a WhP. In sentences where there is no
gap because the WhP occurs in-situ, as in an echo question, or in
multiple question, no P-gaps are allowed.

Then there are Optional and Obligatory P-gaps

"Optional P-gaps"

a) follow the real gap

b) primarily occur in untensed adverbial and complement clauses

c) are in almost free variation with unstressed personal pronouns,
which are coreferential with or bound by the filler

"Obligatory P-gaps"

a) precede the real gap

b) primarily occur in gerunds and noun complements

c) can normally not be replaced by a coreferential pronoun without
significant loss of acceptability.

The Obligatory parasitic Gaps obey the Crossover Constraint:

She also gives evidence that P-gap phenomenon is not limited to NP gaps
but extends to PP and AP gaps as well.

She defines the following restriction on parasitic gaps:

A parasitic gap may not be c-commanded by a real gap.

If we assume that the relation between the real gap and the P-gap is some
form of anaphoric linking, and furthermore P-gaps are necessarily
coreferent with the real gap, then P-gaps should be excluded in those
contexts where the anaphora rules of the language assign disjoint

She also discusses the case of comparatives:


It is the updated version of Kiss 1985, which focuses on the role of
Case properties in P-gap distribution and which considers in detail the
P-gap facts in Hungarian and their parallels to those of English.

As has already been mentioned, the relevance of case compatibility was
first noted by Kiss 1985. In Hungarian, when the case of the P-gap is
different from the case of the real gap, ungrammaticality results. There
are cases when the case of the real gap (nom) is different from the case
of the P-gap (acc), but the sentence is saved by the fact that the
antecedent of the real gap acquires acc in the course of the
derivation. If in such sentences the optional process of case change
does not take place, i.e., if the nom complement extracted out of the object
clause is not assigned acc, then the nom operator will not license an acc

Kiss's condition on case identity is the following, generalized to all

1) In a Parasitic gap construction, the syntactic features of both the
real gap and the parasitic gap are properly transmitted to the
phonologically realized operator.

Kiss argues that the P-gap construction involves two chains (or a
fork in A'-chain) that share the same top section, including the
antecedent. Because the features of both the real gap and the parasitic
gap must be properly transmitted to the shared antecedent, it follows that
the parasitic chain and the licensing chain must have matching
features. As forking A'-chains, even locality features are accounted
for: whereas a link of the licensing chain must be subjacent to the next
higher link of the licensing chain, a link of the parasitic chain must be
subjacent to the next higher link of the parasitic chain or to the link
of the licensing chain.

Various theories of P-gaps will agree that P-gaps are licensed by an
operator variable chain at S-structure. This definition, in Hungarian,
allows not only Wh-operators, but also Focus and Quantifiers, which must
prepose to an A'-position at S-structure, to license P-gaps.



Based on Swedish data, Engdahl deals with the relation between P-gaps and
category restrictions, anaphoric elements, referentiality and island

She gives evidence from Swedish to show that that some languages allow
non-NP P-gaps. Eg. , PPs, adverbials and AP gaps are possible in Swedish
just as, if VP fronting in Swedish is accompanied by "do-support",
examples of parasitic tensed VP-gaps is possible. In Swedish, since the
definite singular pronoun is the proform for Predicate Nominals, they
license pg's. Similarly, since Swedish allows a distinct proform for manner
adverbials which can be used both deictically and anaphorically, just
like a definite proform, consequently, Swedish allows parasitic manner

In Swedish, even amount and duration phrases can license P-gaps depending
on the interpretation, i.e., if they have individual interpretation, and the
right proform exists.

Engdahl points out that Karimi (1999) argues that in Persian, only a
Specific object can license a P-gap. Nonspecific object NPs tend to be
placed immediately before the verb, and constitute a semantic unit with
the verb. Specific object exists independently of the event, can enter
into coreference relations and can license a P-gap. In Swedish, a
nonspecific NP also licenses P-gaps when fronted, but the proform used
for these bare NPs is the singular definite neuter pronoun.

Both Cinque and Karimi assume that good antecedents for P-gaps must be
D-linked, but Engdahl points out that Swedish data shows that this cannot
be a universally valid constraint. Rather one needs to looks at what
types of NPs can be fronted or scrambled, in the language in question, as
this seems to determine what type of P-gaps are possible.

Engdahl even disputes the island sensitivity of P-gaps. According to
her, what is an island in one language may not be an island in another.
Eg. , in Swedish, extraction out of Complex NPs such as relative clauses
are not only acceptable, but are quiet productive. Relative clauses do not
act as islands in some contexts in Swedish.

Thus she concludes that PGs are versatile devices, both
crosslinguistically and within a particular language.


Claim: P-gaps are Resumptive pronouns. He compares properties of P-gap
constructions with those of parallel constructions from Moroccan Arabic,
which include an overt pronoun in the position of the P-gap, as well as
with the classic case of resumption.

He gives the following definition of resumptive pronoun:

A pronoun P is resumptive iff there exists an opertor O such that O
directly A'-binds P at S-structure.

THEORETICAL IMPLICATIONS: (Robert D. Levine, Thomas E. Hukari, and
Michael Calcagno)

They give evidence from English to dispute the claim that P-gaps are all
of the category NP and that P-gaps have inherently pronominal properties.

They prove that the distribution of P-gaps is in no sense incongruent
with that of weak indefinite pronouns.

They use the Head driven Phrase Structure Grammar (HPSG) framework to
explain that the categorial status of P-gaps and their fillers is open,
and also the fact that the content of "parasitic" and true gaps are the

They however concede one apparent empirical challenge to this
treatment--the possibility of case mismatch between gaps in P-gap
construction, which however disappears once certain independently
required modification to the HPSG analysis of case assignment are

Their comments:

Although well formed non-NP P-gaps exist, they need more effort to
construct. Relative intractability of non-nominal P-gaps cannot be a
function of the syntax itself. Interaction of well formed but complex
filler/gap linkages with real time processing limitations is at least
partially responsible for this intractability, along the lines
Kluender 1998 suggests is the source for both strong and weak island


Postal considers putative evidence for pronominal characteristics of
P-gaps in a new area, that of their licensing gaps (L-gaps) .

According to Postal:

1) Most of the current treatments fail to preclude non-DP gaps, which
according to him are perhaps non-existent in English.

2) Even among DP positions, various illicit P-gaps are not properly
blocked. Current views of P-gaps do not yield the sort of parallelisms
between passivization and P-gap constraints for various DP positions.

3) There are constraints on P-gap formation in positions corresponding
to reflexives and reciprocals.

4) Most current frameworks do not block P-gaps in positions that permit
ordinary extraction gaps, but are incompatible with Weak definite
pronouns (Wdps) . These are the antipronominal contexts (ACs) of Postal
1993b, 1994b, 1998. However, P-gaps are systematically precluded in ACs.
Such facts can be explained only by analyzing P-gaps as pronominal. That
is, P-gaps need to regarded as invisible Wdps.

Further Lacunae:

5) No P-gap, even in an otherwise licit P-gap position is licit if it's
L-gap is one of a large class of ACs that he cites in his examples.

Earlier Postal 1993b:747 found that .

The L-gap for P-gap cannot be:

i) a Predicate Nominal position

ii) the post-be DP position for a expletive 'there' construction

iii) the color designating DP of a change of color construction.

Two approaches to the L-gap AC property:

Postal calls the "failure of P-gaps and ACs to intersect", as the P-gap
AC property (PACP), and the "failure of L-gap and AC to intersect" as
the L-gap AC property (LACP) . PACP can be directly explained by assuming
6) :

6) Each P-gap position contains an invisible Wdp.

To explain LACP, he introduces:

7) P-gap structures also involve invisible Wdps in the L-gap position.

Thus, the LACP would be imposed exactly as the PACP is.

To put 5) in perspective, Postal 1994b, 1998 claims that English left
extraction of NPs divide into two major types, called there A-extractions
and B-extractions. The idea is that whereas the latter (B-extraction)
requires invisible resumptive Wdps in extraction sites, the former
do not. The basic evidence for this distinction is precisely that
B-extraction sites, unlike A-extraction sites are incompatible with
ACs. The B-extractions recognized are (DP) topicalizations (DP) clefting
and (DP) non restrictive relative extraction.

Postal divides the ACs into two types. Members of Type I ACs determine
both the PACP and LACP. Type II ACs determine only PACP. Third
explaination of the LACP involves the constraints determining Type I ACs.

9) A Type II AC is induced via logical entailment from a grammatical
condition that does not reference the distribution of Wdps.


In the Metagraph grammar (MGG) framework of Johnson & Postal 1980,
Postal (1986, 1989, 1990, 1991, 1992, 1996), a basic hypothesis about the
nature of P-gap structures combines with a specific reconstruction of the
nature of Type I ACs to yield the LACP & PACP for pg structures.

Postal's basic assumption about P-gap structures is that such structures instantiate a kind of Control relation. Here, that relation is
reconstructed as erasure of a specific type of arc, by an overlay arc.

CTRL-arcs have four properties under the MGG theory:

a) A CTRL-arc must be a weak pronominal, which can be taken to mean
that its head node represents a Wdp.

b) Each CTRL-arc must have an erasure.

c) Each CTRL-arc must be anaphorically paired with the arc that
erases it.

d) The erasure indicated in c) is exclusively local.

Postal acknowledges that his proposals can be doubted.



Postal calls the doctrine that P-gaps can only be NPs as the P-Gap NP
Restriction (pg=NP) . In this doctrine, it would be required that the
"licensing gap" associated with the P-gap be an NP, and also the
"extractee" (the extracted phrase-the WhP for example), be an NP. Postal
refers to this requirement as the P-Gap Licensing Restriction (PLR) .

However, he recognizes that there are counterexamples to this doctrine,
viz. ,

Topicalized "that" clauses license P-gaps

Object-raised "that" clauses license P-gaps

Similarly, for some speakers, clefts as "that" clauses can license
P-gaps, just as clefted NPs can.

Authier (1989, 1991) observes that in certain contexts Complex NP shift (CXS)
seems to involve a "that" clause and the shifted clause can apparently
license a P-gap. Similarly, Right Node Raising (RNR) structures involving
"that" clauses license P-gaps.

However, considerable evidence can be found for both the topicalization
and "object raising" constructions, supporting the view that an
invisible (resumptive) pronoun exists in their complements.

What are Pseudo-Parasitic gaps?

Apparent-P-gaps (A-P) gaps are not true P-gaps, they are pseudo-P-gaps, and
in general are not required to meet the conditions on P-gaps such as
Island Condition and pg=NP.

P-gaps represent a Control phenomenon, in the same sense as the well
known equi-constructions with verbs like 'want'. Postal interprets
control as involving a mechanism that determines the non-presence in
surface forms of ordinary definite pronouns. Postal assumes that
universally a pronoun is subject to control only if it is L-extracted.
Thus, as in Chomsky's 1986b "chain composition" proposal about P-gaps
he takes each English P-gap case to involve at least two extractions each
of which is is an L-extraction. Postal then goes on to explain the
phenomenon in terms of 'arc erasure', and to explain the advantages of
the 'Control view' over Chomsky's 'chain composition' view.

Postal enumerates the following Property Clustering to distinguish
between Parasitic gaps and Pseudo Parasitic Gaps:


i) A gap is a P-gap only if it's existence is dependent on that of
another gap which is a true extraction gap.


A P-gap "licensed" by a gap G cannot occur internal to an island E not
containing G unless E is coextensive with the entire parasitic domain.

C) pg=NP

A P-gap is an NP


The "licensing" category (the extractee from the position of the true
gap) of a P-gap is an NP.


P-gaps cannot occur in positions incompatible with definite pronouns.


If an NP alternates with a P-gap, then the NP must not be inherently be


Neither a P-gap nor its "licensing" category can be a "predicate

Among all gaps satisfying the dependency condition, there are two
distinct classes:

The A-gaps satisfy all the conditions B) to G) above in addition to the
condition that they are "licensed" by L-extraction.

But B-gaps do not satisfy any of these conditions. They designate those
dependent gaps "licensed" by either by CXS or by RNR, i.e., by
R-extraction. The B-gaps are the Pseudo-P-gaps.

He characterizes Pseudo P-gaps as:

1) In all pseudo-P-gap structures, both the "licensing" gaps and
the pseudo-P-gaps are Right Node Raising (RNR) structures.

RNR is a type of R-extraction, and that standard coordinate occurrences of
it are instances of Across-the-Board (ATB) R-extraction.

Properties of RNR:

1) It is insensitive to those island constraints called "subjacency".

2) English coordinate RNR is not limited to NPs but takes T's as

3) Coordinate RNR is not subject to pronominalization conditions-the gaps
preclude pronouns

4) Passivization Constraint:

Insensitivity to Passivization constraint holds for coordinate and
non-coordinate RNR.

5) Predicate Nominal Condition does not hold for coordinate or
non-coordinate RNR.


What was previously taken to be an essentially homogenous P-gap domain
divides into two separate phenomenon. Those A-P gaps "licensed" by
L-extraction other than those of Parenthetical cases is "true P-gaps"
and are a 'Control Phenomenon'. But those A-P gaps "licensed" by
R-extractions instantiate a 'non-control' phenomenon, which is analyzed
as a special case of non-coordinate RNR.

It is also argued that A-P gaps "licensed" by the extractions linked to
parentheticals are not P-gaps but RNR cases this property being masked by
the fact that parentheticals also involve "secondary" L-extractions. The
latter determine that these A-P gaps, unlike those only linked to
R-extractions obey "subjacency" requirements.

 (Andreas Kathol)

Kathol argues that standard German lacks true P-gaps. He takes the
claimed P-gaps in German to instantiate the distinct pseudo P-gap
phenomenon, discussed in Postal 1994a, (which is reprinted in Chapter 8
above) . The latter paper questions the existence of P-gaps whose licensing
gaps involve right extractions. It proposes to treat a range of putative
P-gaps as being instead right node-raising gaps.

A number of properties were identified that group these constructions
with English pseudo-P-gap constructions, in particular the
"non-coordinate Left Node raising (LNR) ", as initially suggested by
Postal 1994a:113. However, these constructions cannot be explained in
their entirety in these terms. The lack of any long distance dependencies
and the sensitivity to the choice of prepositions suggests that the
phenomenon is lexically mediated by the prepositional head of the
adverbial phrase, which in turn is responsible for the valence
requirements of the verbal predicates involved.


 (Christine Tellier)

In her paper in this volume, Tellier discusses three main differences
between English and French P-gaps, all of which have to do with the
domain of the P-gap. Unlike in English, French disallows P-gaps in
adjuncts and relative clauses. And French permits P-gaps in definite NPs
whereas English does not.

P-gaps within tensed adjuncts are totally excluded in French, though
there is a relatively a lack of tense effect in English. At the same
time, with respect to tensed Wh-island violations, French patterns like
English, i.e., it does not allow extraction across Wh-island.

How to account for this difference between French (more generally Romance)
and English?

If pg's are derived by the movement of a null operator to Spec CP of the
adjunct clause (Contreras 1984, Chomsky 1986b), and if the null operator
must be linked to an overt antecedent by chain composition, to be
"strongly bound" in the sense of Chomsky (1986a:85), and if chain
composition must obey the rules of subjacency, then also the difference
between English and French cannot be accounted for, because, as has been
pointed out above, French patterns together with English with respect to
tensed Wh-island.

Tellier attributes the observed differences to the differences in the
agreement properties of the functional categories C and D in the two


Munn treats the relation between P-gaps and logical concepts such as
individual variable as well as the link between P-gaps and coordination
and pronouns.

Munn shows that ATB gaps are similar to adjunct P-gaps with respect to
crossover effects, particularly WCO. The lack of WCO effects in noninitial
conjuncts of an ATB extraction are plausibly attributable to a
resumptive-pronoun strategy within the noninitial conjunct. ATB gaps
however, do not show the category restrictions that P-gaps do, and thus
must also be able to allow "real" (i.e., nonresumptive) extraction in
their noninitial conjuncts. If the category restrictions also follow from
the resumptive pronoun strategy, then ATB gaps must only use this
strategy optionally, whereas P-gaps must use it obligatorily. Munn tries
to reduce the category restrictions to the semantic distinctions between
individual-denoting variables and variables of higher semantic type. Null
resumptive pronouns are only able to refer to individuals. The obligatory
use of the resumptive strategy in the parasitic domain but not in the ATB
extraction then follows naturally from the individual differences between
the two types of domains: real extraction of a non-individual denoting
null operator in the adjunct parasitic domain will always be blocked by
the presence of an intervening operator independently needed by the
semantics of the adjunct clause. Such operators are not present in the
coordinate structures and thus ATB extraction does not show category
restrictions. The question of NP domain parasitic gaps remain open


This is Kennedy's 1997 paper where he argues that contrary to an earlier
claim, a certain class of cases do not involve P-gaps "inside" elided
VPs. Instead they are ordinary pronouns, and are an instance of what
Fiengo & May 1994 call "vehicle change".

Kim and Lyle (1995, 1996) observe that apparent parasitic gaps
constructions do not show island effects when the gap is contained in a
deleted VP. It is a well known fact that parasitic gaps are sensitive to
islands, as a) below shows, but when the gap is enclosed in a deleted VP,
the island effect disappears, be they the case of wh-islands, extraction
from an adjunct, complex NP, coordinate structure constraint etc.

1a) *which article1 did you read t1 after Jim asked [who would be willing
to summarize pg1]?

1b) which article1 did you read t1 after Jim asked [who would be willing
to ------]?

Kennedy argues that cases like b) above do not show Subjacency effects
because they do not contain parasitic gap chains at all. According to
Kennedy, the gap in the deleted VP is a pronoun, which is interpreted as
a variable bound by the wh-operator. He argues that the actual LF
representation of the b) above is:

1c) which article1 did you read t1 after Jim asked [who would be willing
to read it1]?

1c) is perfect.

Kennedy shows that the "gap" in the deleted VP cases is not a sensitive
to Strong Crossover (Condition C), which would be the case if it were a
true parasitic gap, but is sensitive to Condition B, which it would be,
if it is a pronoun.


Postal extends Kennedy's argumentation and argues that although
Kennedy's claim is essentially correct for the class of cases he
treated, application of the same and related argumentation to a different
range of data supports the claim that English has P-gaps "inside" of
verbal ellipsis structures.


The editors claim that to better understand the current state of the
topic, minimally six elements are needed.

1) One needs to have a good overall view of both the documented
factual uniformity and diversity of P-gaps.

2) It is important to establish an idea of the boundary conditions
that divide P-gaps from non-P-gaps, which have been taken to be P-gaps on

3) It is necessary to grasp the relation that exists between P-gaps
and the diverse theoretical proposals that have been claimed either to
illuminate their nature or to be supported/disconfirmed by P-gap

4) One should establish clear generalizations as to how P-gaps relate
to a variety of other syntactic properties including grammatical
category, grammatical agreement, word order, command relations,
grammatical case, distinctions between extraction gaps and other types of
gaps, differences between pronominal and non-pronominal forms, control
constructions, conditions on extraction (e.g., islands), condition on the
distribution of pronominal forms, and so on. .

5) One should understand the history of the research on P-gaps

6) One should have available as complete an account as possible of
the literature on the topic.

I agree with the editors that the present volume fulfils all these
conditions. There are however, some comments which need to be made
especially regarding the points 1), 3) & 4) above.

A) Since it is an assortment of papers, the framework on which the
solution to the problems related to pg's has been sought is different.
For example, Postal's paper proposes a solution in MGG framework, Levine,
Hukari and Calcagno's paper proposes a solution in HPSG
framework. A complete solution to the problem can only be found when
P-gaps are not seen as a phenomenon in isolation but as a phenomenon
intertwined with all other phenomenon of the language, and also with the
entire theoretical framework one is pursuing.

B) Not much work seems to have been done on so called Scrambling
languages, in particular on Indian languages. As pointed out in point
no. 6) above, the editors have certainly tried to present as comprehensive
an account as possible of the literature on the topic. So, if not much
work on Indian languages has been included in the issue, it must be
because not much work has been done on the topic.

However, why it is so important to include that in any work on P-gaps
will become clear from the points I mention below.

In Banerji 2001, in my work on Bangla Syntax in the Mininmalist
framework, I have found that Bangla brings out many issues very
explicitly, which other languages do not e.g.:

1) P-gaps in Bangla can be licensed by Wh-in-situ, by the WhP which has
moved to the TP Phase or by a WhP which has moved to Clause external Spec
CP. That is, what has always been considered a test for A'-position,
i.e., the licensing of a P-gap by a WhP fails in Bangla.

2) A P-gap in Bangla can either be a 'gap', i.e., phonologically null, or
a pronoun (not just a proform) in the real sense of the term. However,
there is a difference in interpretation in the two cases. A P-gap is
licensed as a 'gap' only when the verb is in-situ (SOV order) . But verb
movement from it's in-situ position, to the head of the WhP, or to the
'v' head, causes the P-gap to take a 'pronominal' form. A Parasitic
gap' as a 'gap' is a TYPE operator, but a Parasitic gap as a
'pronominal' is a TOKEN Operator. This is the point which I feel has not
been investigated, in any of the works.

3) PPs in Bangla can license P-gaps, but then, they are better in
'pronominal' form. The pronominal form must however be distinguished
from the 'proform' of Engdahl, because with PPs, they occur as a
complement of P, taking the place of the noun, and not as a proform for
the PP. And since the P-gap in case of PPs are in pronominal form, the
verb must move up, preferably to the Wh-head, in the matrix clause.

(The fact that many languages do not allow PPs as antecedent of P-gaps
must be linked to the fact that they allow P-stranding--something that
Bangla allows with difficulty) .

4) Non-referential phrases cannot license P-gaps, and only Specifics
 (NP/PP) can license PPs as in case of Persian noted by Karimi.

5) The case compatibility is taken care of in the sense that a 'null
P-gap' is allowed only under case compatibility, but it doesn't seem to
be necessary when the P-gap is an overt pronoun or a PP with a pronominal
complement. Conversely, when the case compatibility does not exist, The
P-gap must be expressed overtly.

5) The most important test of a P-gap being in A'-position, and also
being a 'true P-gap' is the fact that it shows 'island
constraints'. It is also a test whether a P-gap is a true gap or a Pseudo
Gap in terms of Postal . Bangla P-gaps, do show island conditions--whether
they are in overt or non-overt form. However, whereas these island
conditions can be voided under certain conditions for the 'pronominal'
forms, (including PPs), they cannot be voided when the P-gap is
"phonologically null". Thus, neither the distinction between true gaps
and Pseudo gaps is that straight forward, nor is the A/A'-distinction.

What I would like to emphasize is that, all the CCPs that Culicover has
enumerated in his paper are in turn linked to larger issues, and they
themselves are not the issues as such. This is what Engdahl in Chapter 4

7) Similarly, the anti-c-command condition for P-gaps is a non-issue,
because it is linked to wider issues, e.g., to the clause structure of
Bangla, and to that of pronominal binding.

8) Finally, I would like to point out that it has been proved extensively
in Banerji 2001, that what has been believed to be scrambling
(Bangla--Sengupta 1990, Hindi-Mahajan 1990, Kidwai 1995), is actually
movement to Clause internal Spec Top/Foc/WhP or to Clause external Spec
Top/CP. That C-system can be a parameter is a point widely ignored in the
literature on Scrambling. In Bangla, the FocP is placed in the TP Phase,
between TP and AgroP, and just as CP and TopP are placed outside TP, as
is customary in majority of the languages, there is another set of TopP
and WhP in the TP Phase. Given this, and from what has been pointed out
above, it is easy to see that A/A'-distinction in the traditional sense
becomes quite meaningless.

Thus, though it is an excellent and a unique book on the subject of
parasitic gaps, very readable too, in spite of some repetitions, it is not
an end in itself. It opens up larger vistas for research, and at the same
time making research so much easier by bringing all the relevant
literature under one roof.


Baker, M. 1995. The Polysynthesis Parameter. Oxford: Oxford University

Banerji, S. 2001. Bangla Syntax-Overtly Minimalist. Revised version of
the PhD thesis submitted to CALTS, University Of Hyderabad, India. Work
in progress.

Chomsky, N. 1981. Lectures on government and binding. Dordrecht: Foris.

Chomsky, N. 1982. Some concepts and consequences of the theory of
government and binding. Cambridge, Mass. : MIT Press.

Chomsky, N. 1986a. Barriers. Cambridge, Mass. : MIT Press.

Chomsky, N. 1986b. Knowledge of language. New York; Praeger.

Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass: MIT Press.

Chomsky, N. 1998. Minimalist Inquiries: The Framework. MIT Occasional
Papers in Linguistics No. 15. MITWPL.

Chomsky, N. 1999. Derivation By Phase. MIT Occasional Papers in
Linguistics No. 18. MITWPL

Chomsky, N. , and H. Lasnik. 1993. The theory of principles and
parameters. In J. Jacobs, A. von Stechow, W. Sternefeld, and T.
Vennemann, eds., Syntax: An International Handbook of
Contemporary Research. Berlin: de Gruyter.

Cinque, G. 1990. Types of A-bar Dependencies. Cambridge, Mass. : MIT

Diesing, Molly. 1992. Indefinites. MIT press, Cambridge, Mass.

Hornstein, N. 1995. Logical Form from GB to Minimalism. Oxford: Blackwell.

Kayne, R. 1994. The antisymmetry of syntax. Cambridge, Mass. : MIT Press.

Kayne, R. 1998. Overt vs Covert movement. ms. New York University.

Kidwai, A. 1995. Binding and Free word order phenomena in Hindi and
Urdu. PhD dissertation, JNU, New Delhi.

Lasnik, H. , and T. Stowell. 1991. Weakest crossover. Linguistic Inquiry 22,

Mahajan, A. 1990. The A/A-bar distinction and movement theory. Doctoral
dissertation, MIT.

Rizzi, L. 1990. Relativized Minimality. Cambridge, Mass: MIT Press.

Rizzi, L. 1995. The fine structure of the left periphery, ms. , Universite
de Geneve.

Sengupta, G. 1990. Binding and Scrambling in Bangla. Doctoral
dissertation, University of Massachusetts, Amherst.

I have recently submitted my thesis at the Centre for Applied Linguistics
and Translation Studies, University Of Hyderabad, India.
It is a work on Bangla, in the Minimalist framework. Earlier, I did
my MPhil in Linguistics from Jawaharlal Nehru University, New Delhi.
I wish to continue working on Bangla and other Indian languages in the
Minimalist framework. It is my dream to be able to work on a Computational
Linguistics project (perhaps a Lexicon project to start with), which uses
the Minimalist framework.
net">Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue